04.08.2013 Views

Hybrid Higher Order FEM-MoM Analysis of ... - Telfor 2010

Hybrid Higher Order FEM-MoM Analysis of ... - Telfor 2010

Hybrid Higher Order FEM-MoM Analysis of ... - Telfor 2010

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

18th Telecommunications forum TELFOR <strong>2010</strong> Serbia, Belgrade, November 23-25, <strong>2010</strong>. <br />

<br />

Abstract — A novel higher order entire-domain finite<br />

element technique is presented for accurate and efficient fullwave<br />

three-dimensional analysis <strong>of</strong> electromagnetic<br />

structures with continuously inhomogeneous material<br />

regions, using large (up to about two wavelengths on a side)<br />

generalized curved hierarchical curl-conforming hexahedral<br />

vector finite elements (<strong>of</strong> arbitrary geometrical and fieldapproximation<br />

orders) that allow continuous change <strong>of</strong><br />

medium parameters throughout their volumes. The results<br />

demonstrate considerable reductions in both number <strong>of</strong><br />

unknowns and computation time <strong>of</strong> the entire-domain <strong>FEM</strong><br />

modeling <strong>of</strong> continuously inhomogeneous materials over<br />

piecewise homogeneous models.<br />

Keywords — Computer-aided analysis, electromagnetic<br />

analysis, electromagnetic scattering, finite element method,<br />

higher order elements, inhomogeneous media, method <strong>of</strong><br />

moments.<br />

I<br />

<strong>Hybrid</strong> <strong>Higher</strong> <strong>Order</strong> <strong>FEM</strong>-<strong>MoM</strong> <strong>Analysis</strong> <strong>of</strong><br />

Continuously Inhomogeneous Electromagnetic<br />

Scatterers<br />

Milan M. Ili, Member, IEEE, Slobodan V. Savi, Andjelija Ž Ili, Member, IEEE, and<br />

Branislav M. Notaroš, Senior Member, IEEE<br />

I. INTRODUCTION<br />

N electromagnetics (EM), the finite element method<br />

(<strong>FEM</strong>) in its various forms and implementations [1]-[4]<br />

has been effectively used for quite some time in full-wave<br />

three-dimensional (3D) computations based on<br />

discretizing partial differential equations. A tremendous<br />

amount <strong>of</strong> effort has been invested in the research <strong>of</strong> the<br />

<strong>FEM</strong> technique in the past 4 decades, making <strong>FEM</strong><br />

methodologies and techniques extremely powerful and<br />

universal numerical tools for solving a broad range <strong>of</strong> both<br />

closed-region (e.g., waveguide and cavity) and openregion<br />

(e.g., antenna and scattering) problems. In the case<br />

<strong>of</strong> open-region problems, hybrid finite element-boundary<br />

integral (FE-BI) technique is used for the exact truncation<br />

<strong>of</strong> the unbounded spatial domain [5]-[6].<br />

For modeling and analyzing structures that contain<br />

This work was supported by the Serbian Ministry <strong>of</strong> Science and<br />

Technological Development under grant ET-11021 and by the National<br />

Science Foundation under grants ECCS-0647380 and ECCS-0650719.<br />

Milan M. Ili is with the School <strong>of</strong> Electrical Engineering, University<br />

<strong>of</strong> Belgrade, Bulevar kralja Aleksandra 73, 11120 Belgrade, Serbia;<br />

(phone: 381-11-3370101; e-mail: milanilic@etf.rs).<br />

Slobodan V. Savi is with the School <strong>of</strong> Electrical Engineering,<br />

University <strong>of</strong> Belgrade, Bulevar kralja Aleksandra 73, 11120 Belgrade,<br />

Serbia; (e-mail: milanilic@etf.rs).<br />

Andjelija Ž. Ili is with the Vina Institute <strong>of</strong> Nuclear Sciences,<br />

Laboratory <strong>of</strong> Physics 010, P.O. Box 522, 11001 Belgrade, Serbia;<br />

(e-mail: andjelijailic@ieee.org).<br />

Branislav M. Notaroš is with the ECE Department, Colorado State<br />

University, 1373 Campus Delivery, Fort Collins, CO 80523-1373, USA;<br />

(e-mail: notaros@colostate.edu).<br />

843<br />

inhomogeneous and complex electromagnetic materials,<br />

<strong>FEM</strong> technique is very efficient and well established as a<br />

method <strong>of</strong> choice. In almost every <strong>FEM</strong> technique, within<br />

a really abundant and impressive body <strong>of</strong> work in the<br />

field, the theory is developed and <strong>FEM</strong> equations are<br />

derived taking advantage <strong>of</strong> the inherent ability <strong>of</strong> <strong>FEM</strong> to<br />

directly treat continuously inhomogeneous materials<br />

(complex permittivity and permeability <strong>of</strong> the media can<br />

be arbitrary functions, or tensors, <strong>of</strong> spatial coordinates,<br />

e.g., (r)<br />

and (r)<br />

, with r standing for the position<br />

vector <strong>of</strong> a point in the adopted coordinate system).<br />

However, it appears that there are practically no papers on<br />

this subject, presenting technique or computer code that<br />

actually implements (r)<br />

and (r)<br />

as continuous space<br />

function within a finite element (FE), thus enabling direct<br />

computation on finite elements that include arbitrary<br />

(continuously) inhomogeneous material. Instead, <strong>FEM</strong><br />

computations are carried out on piecewise homogeneous<br />

approximate model <strong>of</strong> the inhomogeneous structure, with<br />

(r)<br />

and (r)<br />

replaced by the appropriate piecewise<br />

constant approximations. On the other hand, even from the<br />

geometrical modeling point <strong>of</strong> view, it is much simpler<br />

and faster to generate a model <strong>of</strong> the structure with a<br />

single, or few, large continuously inhomogeneous<br />

elements, than a mesh <strong>of</strong> a graded layered structure. In<br />

most <strong>of</strong> the practical situations, such an approach can<br />

dramatically reduce the time needed for an<br />

electromagnetic modeler to set up the problem and<br />

initially model the geometry, before any mesher [7] can be<br />

used to preprocess the data for analysis.<br />

Numerical modeling employing continuously<br />

inhomogeneous finite elements is expected to find<br />

practical applications in analysis <strong>of</strong> a broad range <strong>of</strong><br />

devices, systems, and phenomena in electromagnetics,<br />

including electromagnetic interaction with biological<br />

tissues and materials, absorbing coatings for reduction <strong>of</strong><br />

radar cross sections <strong>of</strong> targets, scattering and diffraction<br />

from inhomogeneous dielectric lenses used for lens<br />

antennas and related structures.<br />

For fully exploiting modeling flexibility <strong>of</strong> continuously<br />

inhomogeneous finite elements, these elements should be<br />

electrically large, which implies use <strong>of</strong> the higher order<br />

field expansions within the elements, as shown in our<br />

preliminary results in [8]. Since the fields in the low order


<strong>FEM</strong> technique are approximated by the low-order basis<br />

functions, the elements must be electrically very small (on<br />

the order <strong>of</strong> a tenth <strong>of</strong> the wavelength in each dimension).<br />

Subdivision <strong>of</strong> the structure using such elements results in<br />

a discretization <strong>of</strong> the permittivity and permeability<br />

pr<strong>of</strong>iles as well, so that elements can be treated as<br />

homogeneous, i.e. their treatment as inhomogeneous<br />

would practically have no effect on the results. Within the<br />

higher order computational approach [9], on the other<br />

hand, higher order basis functions enable the use <strong>of</strong><br />

electrically large geometrical elements (e.g., on the order<br />

<strong>of</strong> a wavelength in each dimension). We refer to the direct<br />

<strong>FEM</strong> computation on such elements as the entire-domain<br />

or large-domain analysis. Note that, in general, higher<br />

order <strong>FEM</strong> technique [10]-[14] can greatly reduce the<br />

number <strong>of</strong> unknowns for a given (homogeneous or<br />

inhomogeneous) problem and enhance the accuracy and<br />

efficiency <strong>of</strong> the analysis in comparison to the low-order<br />

solutions.<br />

II. THEORY AND NUMERICAL IMPLEMENTATION<br />

Consider an electromagnetic structure that contains<br />

some continuously inhomogeneous material regions, as<br />

shown in Fig. 1. In our analysis method, the computation<br />

domain is first tessellated using higher order geometrical<br />

elements in the form <strong>of</strong> Lagrange-type generalized curved<br />

parametric hexahedra <strong>of</strong> arbitrary geometrical orders<br />

Ku , Kv<br />

, and Kw<br />

( K u,<br />

Kv<br />

, Kw<br />

1<br />

), analytically described<br />

as [10]:<br />

r(<br />

u,<br />

v,<br />

w)<br />

<br />

L<br />

K u<br />

i<br />

K u K v K w<br />

rijk<br />

i0<br />

j0<br />

k 0<br />

( u)<br />

<br />

K<br />

u<br />

<br />

l<br />

0<br />

l<br />

i<br />

L<br />

K u<br />

i<br />

u u l<br />

,<br />

u u<br />

i<br />

l<br />

( u)<br />

L<br />

K v<br />

j<br />

( v)<br />

L<br />

K w<br />

k<br />

( w)<br />

,<br />

1 u,<br />

v,<br />

w 1,<br />

where rijk r(<br />

ui,<br />

v j,<br />

wk<br />

) are the position vectors <strong>of</strong> the<br />

interpolation nodes,<br />

Ku<br />

i<br />

(1)<br />

L represent Lagrange interpolation<br />

polynomials, and similar for L (v)<br />

v<br />

w<br />

j and L k . Equation<br />

(1) defines a mapping from a cubical parent domain to the<br />

generalized hexahedron, as illustrated in Fig. 2.<br />

O<br />

r<br />

r( r)<br />

r( r)<br />

K<br />

n inc<br />

K<br />

H inc<br />

E inc<br />

Fig. 1. Electromagnetic structure with continuously<br />

inhomogeneous material.<br />

The electric field in the element, E ( u,<br />

v,<br />

w)<br />

, is<br />

844<br />

approximated by means <strong>of</strong> curl-conforming hierarchical<br />

polynomial vector basis functions given in [10]; let us<br />

denote the functions by f ( u,<br />

v,<br />

w)<br />

, and the respective<br />

arbitrary field-approximation orders <strong>of</strong> the polynomial by<br />

Nu, , Nv<br />

, and N w ( N u,<br />

Nv<br />

, Nw<br />

1 ). The higher order<br />

hierarchical basis functions with improved orthogonality<br />

and conditioning properties constructed from Legendre<br />

polynomials [12] may also be implemented.<br />

v = 1<br />

u =1<br />

u<br />

w =1<br />

w = 1<br />

w<br />

O'<br />

u = 1<br />

x<br />

v =1<br />

z<br />

O<br />

v<br />

y<br />

v = 1<br />

u =1<br />

u = 1<br />

w =1<br />

uvw ,, )<br />

uvw ,, )<br />

r<br />

r<br />

w = 1<br />

v =1<br />

Fig. 2. Generalized curved parametric hexahedron<br />

defined by (1), with the continuous variation <strong>of</strong> medium<br />

parameters given by (2); cubical parent domain is also<br />

shown.<br />

Continuous variation <strong>of</strong> medium parameters in the<br />

computation model can be implemented in different ways.<br />

We choose to take full advantage <strong>of</strong> the already developed<br />

Lagrange interpolating scheme for defining element<br />

spatial coordinates in (1), which can be conveniently<br />

reused to govern the change <strong>of</strong> both the complex relative<br />

permittivity and permeability, r and r , within the<br />

element shown in Fig. 2, as follows:<br />

K<br />

K<br />

K<br />

u v w<br />

K u K v K w<br />

r ( u,<br />

v,<br />

w)<br />

<br />

r, ijk Li<br />

( u)<br />

L j ( v)<br />

Lk<br />

( w)<br />

, (2)<br />

i0<br />

j 0 k 0<br />

1 u,<br />

v,<br />

w 1,<br />

where r, ijk r<br />

( ui<br />

, v j , wk<br />

) are the relative permittivity<br />

values at the point defined by position vectors <strong>of</strong> spatial<br />

interpolation nodes, r ijk , and similarly for r . With such<br />

representation <strong>of</strong> material properties, we then solve for the<br />

unknown field coefficients by substituting the field<br />

expansion E ( u,<br />

v,<br />

w)<br />

in the curl-curl electric-field vector<br />

wave equation [10], reading:<br />

1<br />

r ( u,<br />

v,<br />

w)<br />

E ( u,<br />

v,<br />

w)<br />

(3)<br />

2<br />

k0<br />

r ( u,<br />

v,<br />

w)<br />

E(<br />

u,<br />

v,<br />

w)<br />

0,<br />

where k 0 00<br />

stands for the free-space wave<br />

number. A standard Galerkin weak-form discretization <strong>of</strong><br />

(3) yields:<br />

<br />

V<br />

<br />

1<br />

r<br />

k<br />

( u,<br />

v,<br />

w)<br />

[ f ( u,<br />

v,<br />

w)]<br />

[<br />

E(<br />

u,<br />

v,<br />

w)]<br />

dV<br />

2<br />

0<br />

<br />

V<br />

<br />

( u,<br />

v,<br />

w)<br />

f ( u,<br />

v,<br />

w)<br />

E(<br />

u,<br />

v,<br />

w)<br />

dV<br />

r<br />

<br />

S<br />

<br />

1<br />

r<br />

t<br />

t<br />

( u,<br />

v,<br />

w)<br />

f ( u,<br />

v,<br />

w)<br />

n [<br />

E(<br />

u,<br />

v,<br />

w)]<br />

dS<br />

,<br />

t<br />

(4)


where V is the volume <strong>of</strong> the element, bounded by the<br />

surface S, n is the outward unit normal on S, and f t are<br />

the testing functions (the same as the basis functions).<br />

Once the field coefficients are found, all quantities <strong>of</strong><br />

interest for the analysis are obtained in a straightforward<br />

manner.<br />

III. RESULTS AND DISCUSSION<br />

As an example <strong>of</strong> entire-domain <strong>FEM</strong> analysis <strong>of</strong> an<br />

open-region continuously inhomogeneous structure,<br />

consider a lossless cubical dielectric ( r 1)<br />

scatterer, <strong>of</strong><br />

the side length 2a, and a linear variation <strong>of</strong> r from r<br />

1<br />

at the surface to r 6 at the center <strong>of</strong> the cube, as shown<br />

in Fig. 3. The scatterer is situated in free space and<br />

illuminated by a uniform plane wave incident normal to<br />

one face <strong>of</strong> the scatterer, as shown. The <strong>FEM</strong> domain is<br />

truncated at the cube faces by means <strong>of</strong> unknown electric<br />

and magnetic surface currents <strong>of</strong> densities J S and J mS ,<br />

respectively, that are evaluated by the <strong>MoM</strong>/SIE, giving<br />

rise to a hybrid higher order <strong>FEM</strong>-<strong>MoM</strong> solution [6].<br />

2a<br />

o o<br />

2a<br />

<strong>FEM</strong><br />

r<br />

6<br />

1<br />

O<br />

2a<br />

J , J S mS<br />

y<br />

x<br />

n inc<br />

<strong>MoM</strong><br />

H inc<br />

Fig. 3. <strong>FEM</strong>-<strong>MoM</strong> analysis <strong>of</strong> a lossless continuously<br />

inhomogeneous cubical dielectric scatterer. Single-element<br />

<strong>FEM</strong> domain with linear variation <strong>of</strong> permittivity.<br />

To represent this permittivity variation using the<br />

expansions in (2), the cube is modeled by 7 trilinear<br />

hexahedral finite elements <strong>of</strong> the first geometrical order.<br />

a<br />

Namely, one small cube-like hexahedral, in length, is<br />

10<br />

situated at the cube center and surrounded by 6 “cushion”like<br />

hexahedra between the central cube and the scatterer<br />

surface, onto witch 6 bilinear quadrilateral <strong>MoM</strong> patches<br />

are attached. The obtained model <strong>of</strong> the scatterer is shown<br />

in Fig. 4. The field/current approximation orders are 5 in<br />

all directions for all FE “cushions” and 4 in all directions<br />

for the central finite element and all <strong>MoM</strong> patches,<br />

resulting in 2560 <strong>FEM</strong> and 354 <strong>MoM</strong> unknowns, and a<br />

total <strong>of</strong> 339 s <strong>of</strong> simulation time for 35 frequencies.<br />

E inc<br />

<strong>MoM</strong><br />

patches<br />

o o<br />

a/10<br />

5.875<br />

5.75<br />

1<br />

0<br />

2a<br />

r<br />

<strong>FEM</strong><br />

elements<br />

x<br />

<strong>MoM</strong><br />

patches<br />

Fig. 4. Model <strong>of</strong> a lossless, continuously<br />

inhomogeneous cubical dielectric scatterer consisting <strong>of</strong> 7<br />

finite elements and 6 quadrilateral <strong>MoM</strong> patches.<br />

To both validate the continuously inhomogeneous<br />

<strong>FEM</strong>-<strong>MoM</strong> model <strong>of</strong> the scatterer and evaluate its<br />

efficiency against the piecewise homogeneous<br />

approximate model, the scattering results <strong>of</strong> the higher<br />

order <strong>FEM</strong>-<strong>MoM</strong> analysis using large finite elements with<br />

continuously changing r are compared with the solution<br />

obtained by higher order <strong>FEM</strong>-<strong>MoM</strong> simulations <strong>of</strong><br />

piecewise homogeneous approximate model <strong>of</strong> the<br />

structure in Fig. 4. Each <strong>of</strong> the 6 “cushions” <strong>of</strong> the<br />

continuous model are replaced, respectively by,<br />

Nl 2,<br />

3,<br />

4,<br />

and 7 homogeneous thin “cushions” (platelike<br />

layers), approximating the continuously<br />

inhomogeneous pr<strong>of</strong>ile, which is illustrate in Fig. 5<br />

for N l 4 . Each plate-like layer in layered models is<br />

represented by 6 finite elements <strong>of</strong> the first geometrical<br />

order. Field-approximation orders in these elements are 2<br />

in the radial direction and 5 in transversal directions.<br />

r<br />

845<br />

6.0<br />

5.5<br />

5.0<br />

4.5<br />

4.0<br />

3.5<br />

3.0<br />

2.5<br />

2.0<br />

r continuous<br />

r piecewise constant<br />

approximation<br />

Fig. 5. Piecewise homogeneous approximate graded model<br />

( N l 4 ) <strong>of</strong> the structure in Fig. 3, and piecewise constant<br />

approximation <strong>of</strong> relative permittivity pr<strong>of</strong>iles.<br />

Shown in Fig. 6 is the normalized (to 2<br />

0 ) monostatic<br />

radar cross section (RCS) <strong>of</strong> the cube (as a function <strong>of</strong><br />

/ being the free space wavelength. We observe<br />

a 0 ), 0<br />

2a<br />

5.875<br />

O<br />

5.156<br />

3.969<br />

2.781<br />

x<br />

1.5<br />

1.0<br />

1.594<br />

r<br />

0.0 0.1 0.2 0.3 0.4 0.5<br />

x/a<br />

0.6 0.7 0.8 0.9 1.0


a very good convergence <strong>of</strong> the results obtained by the<br />

layered <strong>FEM</strong>-<strong>MoM</strong> technique to those for the continuous<br />

<strong>FEM</strong>-<strong>MoM</strong> model, as N l increases, as well as an<br />

excellent agreement between the 7-layer and continuous<br />

<strong>FEM</strong>-<strong>MoM</strong> solution. Theoretically, only an infinite<br />

number <strong>of</strong> layers would give the exact solution to the<br />

problem in Fig. 3.<br />

As an additional verification <strong>of</strong> the analysis, the higher<br />

order <strong>MoM</strong> technique based on the surface integral<br />

equation (SIE), i.e. <strong>MoM</strong>/SIE technique [14], is used to<br />

obtain a reference solution for the 7-layer model. An<br />

excellent agreement <strong>of</strong> the higher order <strong>MoM</strong>/SIE and the<br />

corresponding <strong>FEM</strong>-<strong>MoM</strong> results, for the 7-layer model,<br />

is observed.<br />

The additional data shown in the legend <strong>of</strong> Fig. 6 gives<br />

a comparison <strong>of</strong> the number <strong>of</strong> unknowns used and the<br />

computation time for all 6 models. We conclude from the<br />

data shown that the continuous material model is<br />

substantially more efficient than the layered analysis. It is<br />

34.16 times faster than the most accurate layered solution<br />

(for N l 7 ), and the total number <strong>of</strong> unknowns is<br />

reduced 5.6 times.<br />

2<br />

RCS/ 0<br />

2.0<br />

1.8<br />

1.6<br />

1.4<br />

1.2<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Unknowns Time[s]<br />

Continuosu <strong>FEM</strong>-<strong>MoM</strong><br />

2560-384 339.1<br />

Layered <strong>FEM</strong>-<strong>MoM</strong><br />

2 layers 4820-384 1420.6<br />

3 layers 7080-384 3398.9<br />

4 layers 9340-384 4941.2<br />

7 layers 16120-384 11584.9<br />

Layered <strong>MoM</strong> (Reference)<br />

7 layers 2688 1125.0<br />

0.0<br />

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55<br />

a/ 0<br />

Fig. 6. Normalized monostatic radar cross section <strong>of</strong> the<br />

cubical scatterer in Fig. 3 (0 is the free-space<br />

wavelength). Results using the continuous <strong>FEM</strong>-<strong>MoM</strong>,<br />

four different layered <strong>FEM</strong>-<strong>MoM</strong> models, and reference<br />

<strong>MoM</strong>/SIE model are compared.<br />

All numerical results are obtained using HP EliteBook<br />

8440p notebook computer with Intel i5-540 CPU running<br />

at 2.53 GHz and with 2 GB <strong>of</strong> RAM under Micros<strong>of</strong>t<br />

Windows 7 operating system.<br />

IV. CONCLUSION<br />

A novel higher order entire domain hybrid <strong>FEM</strong>-<strong>MoM</strong><br />

technique for accurate and efficient full-wave 3D EM<br />

analysis using large (up to two wavelengths on a side)<br />

finite elements that allow continuous change <strong>of</strong> medium<br />

parameters throughout their volumes, has been presented.<br />

Lagrange-type generalized curved parametric hexahedra<br />

<strong>of</strong> arbitrary geometrical orders with the curl-conforming<br />

846<br />

hierarchical polynomial vector basis functions <strong>of</strong> arbitrary<br />

field-approximation orders and Lagrange interpolation<br />

scheme for variations <strong>of</strong> medium parameters have been<br />

used.<br />

The validity, accuracy, and efficiency <strong>of</strong> the new<br />

technique have been demonstrated through an example <strong>of</strong><br />

EM scatterer with linearly varying permittivity <strong>of</strong> the<br />

dielectric. The example has shown that effective higher<br />

order <strong>FEM</strong> hexahedral meshes, constructed from a very<br />

small number <strong>of</strong> large finite elements with p-refined field<br />

distributions <strong>of</strong> high approximation orders, is the method<br />

<strong>of</strong> choice for EM modeling <strong>of</strong> structures including<br />

material inhomogeneities. High efficiency, and<br />

considerable reductions in both number <strong>of</strong> unknowns and<br />

computation time <strong>of</strong> the entire-domain <strong>FEM</strong> modeling <strong>of</strong><br />

continuously inhomogeneous material, in comparison with<br />

the piecewise homogeneous (layered) models, have been<br />

demonstrated.<br />

[1]<br />

REFERENCES<br />

P. P. Silvester and R. L. Ferrari, Finite Elements for Electrical<br />

Engineers, 3rd ed., Cambridge University Press, 1996.<br />

[2] J. M. Jin, The Finite Element Method in Electromagnetics, 2nd ed.,<br />

John Wiley & Sons, New York, 2002.<br />

[3] J. L. Volakis, A. Chatterjee, and L. C. Kempel, “Finite Element<br />

Method for Electromagnetics”, IEEE Press, New York, 1998.<br />

[4] J. L. Volakis, K. Sertel, and B. C. Usner, Frequency Domain<br />

<strong>Hybrid</strong> Finite Element Methods in Electromagnetics, Morgan &<br />

Claypool Publishers, 2006.<br />

[5] J.-M. Jin, J. L. Volakis, and J. D. Collins, “A finite-elementboundary-integral<br />

method for scattering and radiation by two- and<br />

three-dimensional structures”, IEEE Antennas and Propagation<br />

Magazine, Vol. 33, pp. 22-32, March 1991.<br />

[6] M. M. Ili, M. Djordjevi, A. Ž. Ili, and B. M. Notaroš, “<strong>Higher</strong><br />

<strong>Order</strong> <strong>Hybrid</strong> <strong>FEM</strong>-<strong>MoM</strong> Technique for <strong>Analysis</strong> <strong>of</strong> Antennas and<br />

Scatterers”, IEEE Transactions on Antennas and Propagation,<br />

August 2008.<br />

[7] S. V. Savi, M. M. Ili, B. M. Kolundžija, and B. M. Notaroš,<br />

“Efficient Modeling <strong>of</strong> Complex Electromagnetic Structures Based<br />

on the Novel Algorithm for Spatial Segmentation Using Hexahedral<br />

Finite Elements”, submitted for <strong>Telfor</strong> Journal, <strong>2010</strong>.<br />

[8] M. M. Ili, A. Ž. Ili, and B. M. Notaroš, “Continuously<br />

Inhomogeneous <strong>Higher</strong> <strong>Order</strong> Finite Elements for 3-D<br />

[9]<br />

Electromagnetic <strong>Analysis</strong>”, IEEE Transaction on Antennas and<br />

Propagation, Vol. 57, No. 9, pp. 2798-2802, September 2009.<br />

B. M. Notaroš, “<strong>Higher</strong> <strong>Order</strong> Frequency-Domain Computational<br />

Electromagnetics”, Special Issue on Large and Multiscale<br />

Computational Electromagnetics, IEEE Transactions on Antennas<br />

and Propagation, Vol. 56, No. 8, pp. 2251-2276, August 2008.<br />

[10] M. M. Ili and B. M. Notaroš, “<strong>Higher</strong> order hierarchical curved<br />

hexahedral vector finite elements for electromagnetic modeling”,<br />

IEEE Transactions on Microwave Theory and Techniques, Vol. 51,<br />

No. 3, pp. 1026-1033, March 2003.<br />

[11] M. M. Ili, A. Ž. Ili, and B. M. Notaroš, “<strong>Higher</strong> order<br />

large-domain <strong>FEM</strong> modeling <strong>of</strong> 3-D multiport waveguide structures<br />

with arbitrary discontinuities”, IEEE Transactions on Microwave<br />

Theory and Techniques, Vol. 52, No. 6, pp. 1608-1614, June 2004.<br />

[12] M. M. Ili and B. M. Notaroš, “<strong>Higher</strong> order large-domain<br />

hierarchical <strong>FEM</strong> technique for electromagnetic modeling using<br />

Legendre basis functions on generalized hexahedra",<br />

Electromagnetics, Vol. 26, pp. 517-529, October 2006.<br />

[13] A. Ž. Ili, S. V. Savi, M. M. Ili, and B. M. Notaroš, “<strong>Analysis</strong> <strong>of</strong><br />

Electromagnetic Scatterers Using <strong>Hybrid</strong> <strong>Higher</strong> <strong>Order</strong> <strong>FEM</strong>-<strong>MoM</strong><br />

Technique”, <strong>Telfor</strong> Journal, Vol. 1, No. 2, pp. 53-56, 2009.<br />

[14] B. M. Kolundžija, J. S. Ognjanovi, T. K. Sarkar, D. S. Šumi, M.<br />

M. Paramenti, B. B. Jani, D. I. Olan, D. V. Toši, M. S. Tasi,<br />

WIPL-D Microwave: Circuits and 3D EM Simulation <strong>of</strong> RF &<br />

Microwave Applications, WIPL-D & Artech House, 2005.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!